JP2017191987A - Echo canceling apparatus, method thereof, program, and recording medium - Google Patents

Abstract

An echo canceller and the like that can reduce residual echoes in an initial state with fewer adaptive filters than in the prior art. An echo canceller includes a multiplier that multiplies a sub-microphone output signal by a coefficient corresponding to a distance d1 from a speaker to a sub-microphone and a distance d2 from the speaker to a main microphone to obtain an estimated value of an echo component; And a subtractor that subtracts the estimated value of the echo component from the output signal of the main microphone to obtain an error signal. [Selection] Figure 2

Description

  The present invention relates to a technique for erasing sound (acoustic echo) that circulates from a speaker to a microphone in a hands-free communication device or the like.

  Patent Document 1 is known as a prior art of an echo canceller.

  FIG. 1 is a functional block diagram of a conventional echo canceling device 90 disclosed in Patent Document 1. In FIG. The conventional echo canceling device 90 picks up the sound of the near-end speaker 1 as much as possible and picks up the sound from the speaker 2 as much as possible, and picks up the sound from the speaker 2 as much as possible and the near-end. A sub microphone 4 arranged so as not to pick up the voice of the speaker 1 as much as possible is used. This is realized by a directional microphone facing the speaker direction, a directional microphone facing the speaker direction, or the like. Echoes are canceled using two adaptive filters, a first adaptive filter unit 5 connected to the sub microphone 4 and a second adaptive filter unit 6 to which a reproduction signal of the speaker 2 is input. The first adaptive filter unit 5 generates a first pseudo echo signal by using the output signal of the sub microphone 4 and the error signal that is the output value of the subtraction unit 7. The second adaptive filter unit 6 generates a second pseudo echo signal using the reproduction signal of the speaker 2 and the error signal. The adding unit 8 obtains a third pseudo echo signal using the first pseudo echo signal and the second pseudo echo signal. The subtraction unit 7 subtracts the third pseudo echo signal from the output signal of the main microphone 3 to obtain an error signal.

JP 2011-160429 A

  However, the conventional echo canceller 90 has three problems. First, since it is necessary to use the main microphone 3 that picks up the sound of the near-end speaker 1 as much as possible and does not pick up the sound of the speaker 2 as much as possible, pick up the sound in the direction of the speaker 2 when viewed from the main microphone 3. This is a problem that the voice of the near-end speaker 1 becomes small when the near-end speaker 1 is in that direction. Secondly, since two adaptive filters (first adaptive filter unit 5 and second adaptive filter unit 6) are used, the amount of calculation is large. The third is that the residual echo in the initial state where the adaptive filters of the first adaptive filter unit 5 and the second adaptive filter unit 6 are not learned is large.

  An object of the present invention is to provide an echo canceller, a method, a program, and a recording medium that have fewer adaptive filters than the prior art disclosed in Patent Document 1 and can reduce residual echo in the initial state. .

  In order to solve the above problem, according to one aspect of the present invention, the echo canceller outputs a coefficient according to the distance d1 from the speaker to the sub microphone and the distance d2 from the speaker to the main microphone. A multiplication unit that multiplies the signal to obtain an estimated value of the echo component, and a subtracting unit that subtracts the estimated value of the echo component from the output signal of the main microphone to obtain an error signal.

  In order to solve the above-described problem, according to another aspect of the present invention, an echo canceller calculates a coefficient according to a transfer characteristic from a speaker to a sub microphone and a transfer characteristic from a speaker to a main microphone. A filtering unit that multiplies the output signal to obtain an estimated value of the echo component, and a subtracting unit that subtracts the estimated value of the echo component from the output signal of the main microphone to obtain an error signal.

  In order to solve the above-mentioned problem, according to another aspect of the present invention, an echo canceling method uses a coefficient according to a distance d1 from a speaker to a sub microphone and a distance d2 from a speaker to a main microphone. A multiplication step for multiplying the output signal to obtain an estimated value of the echo component, and a subtraction step for obtaining an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone.

  In order to solve the above-described problem, according to another aspect of the present invention, an echo canceling method uses a coefficient of a sub microphone according to a transfer characteristic from a speaker to a sub microphone and a transfer characteristic from a speaker to a main microphone. A filtering step of multiplying the output signal to obtain an estimated value of the echo component and a subtracting step of obtaining an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone are included.

  According to the present invention, the near-end speaker's voice is not reduced, the number of adaptive filters is smaller than that of the prior art of Patent Document 1, and the residual echo in the initial state can be reduced.

The functional block diagram of the echo cancellation apparatus which concerns on a prior art. The functional block diagram of the echo cancellation apparatus which concerns on 1st embodiment. The figure which shows the example of the processing flow of the echo cancellation apparatus which concerns on 1st embodiment. The functional block diagram of the echo cancellation apparatus which concerns on the modification of 1st embodiment. The functional block diagram of the echo cancellation apparatus which concerns on 3rd embodiment. The figure which shows the example of the processing flow of the echo cancellation apparatus which concerns on 3rd embodiment. The functional block diagram of the echo cancellation apparatus which concerns on 4th embodiment.

  Hereinafter, embodiments of the present invention will be described. In the drawings used for the following description, constituent parts having the same function and steps for performing the same process are denoted by the same reference numerals, and redundant description is omitted.

<Points of first embodiment>
FIG. 2 is a functional block diagram of the echo cancellation apparatus 100 according to the first embodiment, and FIG. 3 shows an example of the processing flow. The echo cancellation apparatus 100 includes a second adaptive filter unit 106, an addition unit 108, a multiplication unit 109, and a subtraction unit 107.

The echo canceling apparatus 100 receives the reproduction signal x (t) from the speaker 102, the output signal y ₂ (t) from the main microphone 103, and the output signal y ₁ (t) from the sub microphone 104 as input. An error signal e (t) in which the echo component is eliminated from the output signal y ₂ (t) is output. Here, t represents a discrete time.

In the present embodiment, two microphones are used: a sub microphone 104 disposed near the speaker 102 and a main microphone 103 disposed farther from the speaker 102 than the sub microphone 104. The sound radiated evenly in free space has a smaller amplitude in inverse proportion to the distance. If the distance from the speaker 102 to the sub microphone 104 is d1, and the distance from the speaker 102 to the main microphone 103 is d2, the amplitude P1 of the sound observed by the sub microphone 104 (the sound reproduced by the speaker 102) and the main The relationship of the amplitude P2 of the sound observed by the microphone 103 (the sound reproduced by the speaker 102) is shown below.
P1 ＝ d2 / d1 ・ P2 (1)
Using this relationship, the sound (direct sound) that directly reaches the main microphone 103 from the speaker 102 is estimated and subtracted from the output signal y ₂ (t) of the main microphone 103. The output signal y ₁ (t) of the sub microphone 104 is multiplied by a fixed coefficient α calculated from the arrangement of the main microphone 103 and the sub microphone 104, and subtracted by the subtractor 107 from the output signal y ₂ (t) of the main microphone 103. To do. As the fixed coefficient α, for example, a value (d1 / d2) obtained by dividing the distance d1 from the speaker 102 to the sub microphone 104 by the distance d2 from the speaker 102 to the main microphone 103 is set. Further, this coefficient may be set so that the echo contained in the error signal e (t), which is the output value of the subtractor 107, is minimized experimentally.

Since an echo component such as a room reverberation component remains in the error signal e (t) that is an output value of the subtraction unit 107, a second adaptive filter unit that receives the reproduction signal x (t) of the speaker 102 as an input. The residual echo is estimated using 106 and subtracted by the subtractor 107 from the output signal y ₂ (t) of the main microphone 103 to be erased. However, the second adaptive filter unit 106 may be omitted.

With such a configuration, a processing with a very small amount of calculation is performed by multiplying the output signal y ₁ (t) of the sub microphone 104 by a fixed coefficient (subtraction coefficient) α and subtracting it from the output signal y ₂ (t) of the main microphone 103. Thus, it is possible to remove the direct sound component of the echo. In addition, since the subtraction coefficient is set with known information in advance, the echo can be removed from the initial state. If the distance from the speaker 102 to the sub microphone 104 is set to be d1 and the distance from the speaker 102 to the main microphone 103 is set to be sufficiently shorter than d2, the subtraction coefficient is significantly smaller than 1, and the near-end speaker The influence on the sound component of 1 is almost eliminated. For example, if a non-directional microphone is used as the main microphone 103, sound can be collected uniformly in all directions. In addition, since the present embodiment is effective for the direct sound component of the echo, it is particularly effective for a small hands-free device in which the distance between the speaker 102 and the main microphone 103 is short. For example, a hands-free device dedicated to hands-free calling may be used, or a smartphone or tablet provided with a sub microphone near the speaker and a main microphone at a position relatively far from the speaker is used as the hands-free device. You can also

<Echo Canceling Device 100 according to First Embodiment>
The processing contents of the echo cancellation apparatus 100 will be described with reference to FIGS.

The echo canceling apparatus 100 receives the reproduction signal x (t) from the speaker 102, the output signal y ₂ (t) from the main microphone 103, and the output signal y ₁ (t) from the sub microphone 104 as input. An error signal e (t) obtained by eliminating the estimated value of the echo component from the output signal y ₂ (t) is output.

<Multiplier 109>
The multiplier 109 receives the output signal y ₁ (t) of the sub microphone, and a coefficient α (for example, α = α) corresponding to the distance d 1 from the speaker 102 to the sub microphone 104 and the distance d 2 from the speaker 102 to the main microphone 103. By multiplying the output signal y ₁ (t) of the sub microphone by d1 / d2), an estimated value αy ₁ (t) of the echo component is obtained (S1) and output.

  Note that the distances d1 and d2 and the amplitudes P1 and P2 satisfy α = d1 / d2 if the relationship of Expression (1) is satisfied. However, actually, since various factors other than the distances d1 and d2 affect the relationship of the expression (1), α = d1 / d2 is not always satisfied. Therefore, the coefficient α may be experimentally set so that the echo component included in the error signal e (t), which is the output value of the subtractor 107, is minimized. In any case, the point that the coefficient α becomes a value according to the distance d1 and the distance d2 remains the same. For example, α = d1 / d2 + γ, and γ is a value that varies depending on factors other than the distances d1 and d2.

<Second Adaptive Filter Unit 106>
The second adaptive filter unit 106 receives the reproduction signal x (t) and the error signal e (t) of the speaker 102, and uses these values to store the residual signal included in the output signal y ₂ (t) of the main microphone 103. An estimated value β (t) of the echo component is obtained (S2) and output.

For example, the estimated value β (t) is obtained by the following equation using the reproduction signal x (t) and a filter coefficient H (t) described later.
β (t) = H (t) ^T X (t) (2)
H (t) = (h (0), h (1), ..., h (L-1)) ^T (3)
X (t) = (x (t), x (t-1), ..., x (t-L + 1)) ^T (4)
However, the superscript T represents transposition, and ^AT represents the transposition of the vector A.

Here, the filter coefficient H (t) is updated in a filter coefficient update unit (not shown) inside the second adaptive filter unit 106. For example, when the NLMS algorithm is used, the filter coefficient H (t) is updated by the following equation.
H (t + 1) = H (t) + aX (t) e (t) / X (t) ^T X (t) (5)
0 <a <2 (6)
Here, a represents the step size of the NLMS algorithm. The method for updating and obtaining the filter coefficient H (t) is not limited to this method, and a conventional method may be used. For example, there are methods described in Patent Document 1 (LMS algorithm, projection algorithm, RLS algorithm, etc.).

<Adding unit 108>
The adder 108 receives the estimated value αy ₁ (t) of the echo component and the estimated value β (t) of the residual echo component and obtains the sum (αy ₁ (t) + β (t)) (S3), Output.

<Subtraction unit 107>
The subtraction unit 107 receives the output signal y ₂ (t) of the main microphone 103 and the sum (αy ₁ (t) + β (t)), and the difference y ₂ (t) − (αy ₁ (t) + β (t)) is obtained (S4) and output as an error signal e (t).

<Effect>
With the above configuration, even when there is a near-end speaker in the direction of the speaker when viewed from the main microphone, the near-end speaker's voice is not reduced, and the number of adaptive filters is the conventional technology of Patent Document 1. The residual echo in the initial state can be reduced. That is, high echo cancellation can be realized in a main microphone having an arbitrary directivity. In this embodiment, the amount of calculation can be set to one adaptive filter. Since the multiplier 109 estimates the echo without using the adaptive filter, it is possible to reduce the residual echo in the initial state before learning the adaptive filter. The configuration of the present embodiment realizes performance improvement particularly in the case of a small hands-free device in which the microphone and the speaker are very close.

<Modification>
In the present embodiment, the second adaptive filter unit 106 may not be provided (see FIG. 4). In this case, the adding unit 108 may not be provided, and the processes S2 and S3 may be omitted. The subtracting unit 107 receives the output signal y ₂ (t) of the main microphone 103 and the estimated value αy ₁ (t) of the echo component, obtains the difference y ₂ (t) −αy ₁ (t) (S4), Output as error signal e (t).

In the present embodiment, the sub microphone 104 is disposed at a position close to the speaker 102, and the main microphone is disposed at a position far from the speaker 102. For example, when d1 = 1 cm and d2 = 10 cm, α is about 0.1, and the influence on the speech component of the near-end speaker 1 (the influence on the output signal y ₂ (t) of the main microphone 103) is almost eliminated. However, the sub microphone 104 may be disposed at a position far from the speaker 102, and the main microphone may be disposed at a position close to the speaker 102. In error signal e (t) = y ₂ (t)-(αy ₁ (t) + β (t)), α is 1 or more, so the phase of error signal e (t) obtained is reversed. The function of canceling echo is not affected. However, since the output signal of the sub microphone 104 is amplified including noise, the arrangement of the first embodiment is preferable. When d1 and d2 are approximately the same value, α is a value around 1, and the error signal e (t) = y ₂ (t, regardless of the magnitude of the output signal y ₂ (t) of the main microphone 103. )-(αy ₁ (t) + β (t)) becomes a value around 0, and the voice of the near-end speaker becomes small. Therefore, it is desirable that the ratio d1 / d2 between d1 and d2 is smaller than the predetermined value p or larger than the predetermined value q. If d1 is approximately twice d2 or d2 is approximately twice d1, it is desirable to satisfy p <0.5, p> 2 since the near-end speaker's voice can be heard.

<Second embodiment>
A description will be given centering on differences from the first embodiment.

  The echo cancellation apparatus 100 includes a second adaptive filter unit 106, an addition unit 108, a multiplication unit 109, a subtraction unit 107, and further includes a delay unit 110 (see FIG. 2).

A function of delaying the output of the sub microphone 104 by the time τ during which the sound travels a distance obtained by subtracting the distance d 1 from the speaker 102 to the main microphone 103 from the distance d 1 from the speaker 102 to the sub microphone 104 is added.
τ = (d2-d1) / v
Where v represents the speed of sound.

<Delay unit 110>
The delay unit 110 receives the output signal y ₁ (t) of the sub microphone 104, delays it by τ = (d2−d1) / v (S9, see FIG. 3), and outputs it. Therefore, the delayed output signal at time t is y ₂ (t−τ).

<Effect>
By setting it as such a structure, the effect similar to 1st embodiment can be acquired. Furthermore, the time difference can also be subtracted after this, and the echo cancellation performance becomes higher.

  In this embodiment, the delay unit 110 is arranged between the sub microphone 104 and the multiplication unit 109. However, the same effect can be obtained by arranging the delay unit 110 between the multiplication unit 109 and the addition unit 108. Can be obtained.

<Third embodiment>
A description will be given centering on differences from the second embodiment.

  FIG. 5 is a functional block diagram of the echo canceller according to the present embodiment, and FIG. 6 shows an example of the processing flow.

  The echo cancellation apparatus 100 includes a second adaptive filter unit 106, an addition unit 108, a fixed filter unit 111, and a subtraction unit 107. That is, the echo cancellation apparatus 100 according to this embodiment has a configuration in which the delay unit 110 and the multiplication unit 109 are replaced with a fixed filter unit 111 that performs filtering using an FIR filter or the like in the second embodiment.

  It is assumed that there is a difference in the frequency characteristics of the direct sound components of echoes reaching the sub microphone 104 and the main microphone 103 due to the sound radiation characteristics of the speaker 102 and the characteristic differences between the main microphone 103 and the sub microphone 104. Differences in delay time difference, amplitude difference, and frequency characteristic between the sub microphone 104 and the main microphone 103 are obtained through experiments and simulations, the characteristic difference is set in the fixed filter unit 111, and the output of the fixed filter unit 111 is set as the main filter. Subtract from the output of the microphone 103.

<Fixed filter unit 111>
Fixed filter unit 111 receives the output signal y ₁ of the sub microphone 104 (t), the main on the output signal y ₁ (t), the transfer characteristic f ₁ from the speaker 102 to the sub-microphone _104, a _ω a speaker 102 The estimated value of the echo component is obtained by multiplying the coefficient α _ω corresponding to the transfer characteristic f _{2, ω} up to the microphone 103 (S111) and output. Note that ω is an index representing frequency, f _{1, ω} represents the transfer characteristic of the sub microphone 104 at the frequency ω, and f _{2, ω} represents the transfer characteristic of the main microphone 103 at the frequency ω. For example, the coefficient α _ω at the frequency ω is set to α _ω = f _{2, ω} / f _{1, ω} .

For example, frequency domain coefficients (α ₁ , α ₂ ,…, α _ω ,…, α _Ω ) are converted to time domain coefficients (α (0), α (1,…, α (Ω-1)). Then, the output signal y ₁ (t), y ₁ (t-1), ..., y ₁ (t-Ω + 1) is multiplied to obtain the sum to obtain the estimated value of the echo component, that is, the echo at time t If the component is b (t),
b (t) = Σα (m) y ₁ (tm)
It becomes.

Also, for example, the time domain output signal y ₁ (t) is converted to the frequency domain output signal y ₁ (1), y ₁ (2), ..., y ₁ (ω), ..., y ₁ (Ω). Multiply frequency domain coefficients (α ₁ , α ₂ ,…, α _ω ,…, α _Ω ) for each frequency ω, and convert the multiplication result (product) y ₁ (ω) α _ω into a time domain signal. May be.

  FFT (short-time Fourier transform) or the like can be used as a method for changing a time-domain value to a frequency-domain value, and a corresponding method (for example, IFFT) can be used to change a frequency-domain value to a time-domain value. Inverse short time Fourier transform) may be used.

It should be noted that the coefficients (α ₁ , α ₂ ,..., Α _ω ,..., Α _Ω ) (or those converted into time domain coefficients) are obtained in advance through experiments and simulations.

<Effect>
By adopting such a configuration, the delay time difference and the amplitude difference difference are incorporated in the coefficient α _ω , so that the same effect as in the second embodiment can be obtained. Further, since the difference in frequency characteristics can be taken into account, the estimation accuracy of the echo direct sound component is improved, and the echo cancellation performance is improved.

<Fourth embodiment>
A description will be given centering on differences from the third embodiment.

  The main microphone 103 of the third embodiment is replaced with M microphones 103-m and a beam forming unit 112 (see FIG. 7). However, m = 1, 2,..., M, and M is an integer of 2 or more.

<M microphones 103-m and beam forming unit 112>
The beam forming unit 112 receives output signals from the M microphones 103-m, performs beam forming processing, obtains and outputs an output signal y ₂ (t) having directivity in a predetermined direction.

In this way, by using the M microphones 103-m and the beam forming unit 112, various directivities can be formed. When such beam forming is combined with the present invention, the delay time difference, amplitude difference, and frequency of the direct sound of the echo between the output signal y ₂ (t) and the output signal y ₁ (t) of the sub microphone 104 The difference in characteristics is estimated in advance by experiment or simulation, and based on this, the filter coefficient of the fixed filter is set as in the third embodiment.

  As a result, the beam forming technique can be used together.

  In addition, the beam forming of this embodiment is a process which filters the output of a microphone microphone with a fixed filter coefficient, and outputs the sum of the outputs.

<Modification>
In the present embodiment, the beam forming unit 112 is a part of the echo canceling device 100, but the echo canceling device 100 may not include the beam forming unit 112, and may receive a signal after the beam forming process.

  You may combine 1st embodiment, 2nd embodiment, its modification, and this embodiment.

<Other variations>
The present invention is not limited to the above-described embodiments and modifications. For example, the various processes described above are not only executed in time series according to the description, but may also be executed in parallel or individually as required by the processing capability of the apparatus that executes the processes. In addition, it can change suitably in the range which does not deviate from the meaning of this invention.

<Program and recording medium>
In addition, various processing functions in each device described in the above embodiments and modifications may be realized by a computer. In that case, the processing contents of the functions that each device should have are described by a program. Then, by executing this program on a computer, various processing functions in each of the above devices are realized on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Further, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its storage unit. When executing the process, this computer reads the program stored in its own storage unit and executes the process according to the read program. As another embodiment of this program, a computer may read a program directly from a portable recording medium and execute processing according to the program. Further, each time a program is transferred from the server computer to the computer, processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program includes information provided for processing by the electronic computer and equivalent to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In addition, although each device is configured by executing a predetermined program on a computer, at least a part of these processing contents may be realized by hardware.

Claims (8)

A multiplier that multiplies the output signal of the sub microphone by a coefficient corresponding to the distance d1 from the speaker to the sub microphone and the distance d2 from the speaker to the main microphone to obtain an estimated value of the echo component;
A subtracting unit that obtains an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone,
Echo canceler. The echo canceller of claim 1,
A delay unit that delays an output value of the sub microphone or an estimated value of the echo component according to a difference between the distance d1 and the distance d2.
Echo canceler. A filtering unit that multiplies the output signal of the sub microphone by a coefficient according to the transfer characteristic from the speaker to the sub microphone and the transfer characteristic from the speaker to the main microphone to obtain an estimated value of the echo component;
A subtracting unit that obtains an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone,
Echo canceler. The echo canceller according to any one of claims 1 to 3,
The main microphone is composed of a plurality of microphones and a beam forming unit,
The beam forming unit performs a beam forming process using output signals of the plurality of microphones to obtain an output signal having directivity in a predetermined direction;
Echo canceler. Multiplication step of multiplying the output signal of the sub microphone by a coefficient according to the distance d1 from the speaker to the sub microphone and the distance d2 from the speaker to the main microphone to obtain an estimated value of the echo component;
Subtracting an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone,
Echo cancellation method. A filtering step of multiplying the output signal of the sub microphone by a coefficient according to the transfer characteristic from the speaker to the sub microphone and the transfer characteristic from the speaker to the main microphone to obtain an estimated value of the echo component;
Subtracting an error signal by subtracting the estimated value of the echo component from the output signal of the main microphone,
Echo cancellation method.   A program for causing a computer to function as the echo canceling apparatus according to claim 1.   A computer-readable recording medium on which a program for causing a computer to function as the echo canceling device according to any one of claims 1 to 4 is recorded.

Images